Reconstructing the emergence of a lethal infectious ...rspb.royalsocietypublishing.org/content/royprsb/283/1839/20160952...Electronic supplementary material is available ... 3School

on July 3, 2018http://rspb.royalsocietypublishing.org/Downloaded from

rspb.royalsocietypublishing.org

ResearchCite this article: Price SJ, Garner TWJ,Cunningham AA, Langton TES, Nichols RA.

2016 Reconstructing the emergence of a lethal

infectious disease of wildlife supports a

key role for spread through translocations

by humans. Proc. R. Soc. B 283: 20160952.http://dx.doi.org/10.1098/rspb.2016.0952

Received: 29 April 2016

Accepted: 31 August 2016

Subject Areas:ecology, health and disease and epidemiology

Keywords:pathogen pollution, ranavirus, citizen science,

wildlife disease, anthropogenic drivers,

spatio-temporal models

Author for correspondence:Stephen J. Price

e-mail: [email protected]

Electronic supplementary material is available

online at https://dx.doi.org/10.6084/m9.fig-

share.c.3469740.

& 2016 The Authors. Published by the Royal Society under the terms of the Creative Commons AttributionLicense http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the originalauthor and source are credited.

Reconstructing the emergence of a lethalinfectious disease of wildlife supports akey role for spread through translocationsby humans

Stephen J. Price1,2,3, Trenton W. J. Garner2, Andrew A. Cunningham2,Tom E. S. Langton4 and Richard A. Nichols3

1UCL Genetics Institute, University College London, Darwin Building, Gower Street, London WC1E 6BT, UK2Institute of Zoology, Zoological Society of London, London NW1 4RY, UK3School of Biological and Chemical Sciences, Queen Mary University of London, Mile End Road,London E1 4NS, UK4Herpetofauna Consultants International, Triton House, Bramfield, Halesworth, Suffolk IP19 9AE, UK

SJP, 0000-0001-6983-6250; TWJG, 0000-0002-0336-9706; RAN, 0000-0002-4801-9312

There have been few reconstructions of wildlife disease emergences, despitetheir extensive impact on biodiversity and human health. This is in large partattributable to the lack of structured and robust spatio-temporal datasets. Weovercame logistical problems of obtaining suitable information by using datafrom a citizen science project and formulating spatio-temporal models of thespread of a wildlife pathogen (genus Ranavirus, infecting amphibians). Weevaluated three main hypotheses for the rapid increase in disease reportsin the UK: that outbreaks were being reported more frequently, that climatechange had altered the interaction between hosts and a previously wide-spread pathogen, and that disease was emerging due to spatial spread ofa novel pathogen. Our analysis characterized localized spread from nearbyponds, consistent with amphibian dispersal, but also revealed a highlysignificant trend for elevated rates of additional outbreaks in localitieswith higher human population densitypointing to human activities inalso spreading the virus. Phylogenetic analyses of pathogen genomessupport the inference of at least two independent introductions into theUK. Together these results point strongly to humans repeatedly translocat-ing ranaviruses into the UK from other countries and between UK ponds,and therefore suggest potential control measures.

1. BackgroundEmerging infectious diseases (EIDs) are defined as diseases undergoing anincrease in incidence, geographical range, or host range. EIDs of humans, live-stock, and crops are increasingly recognized as major challenges, because theycan impose massive economic burdens and have major public health impli-cations [1]. By contrast, much interest in wildlife diseases has been indirect, aconsequence of wildlife populations serving as reservoirs for human diseases(zoonoses, see [2]) and diseases of livestock (e.g. bovine tuberculosis and rinder-pest [3,4]). A second, more direct motivation for understanding wildlife EIDsis their impact on biodiversity, since they can cause extirpation and/orcatastrophic multihost declines [59].

Preventing EIDs at source is highly desirable; but such intervention requiresa thorough understanding of the drivers of emergence. Reconstruction ofmodes of transmission and patterns of spread can inform strategic managementapproaches. For example, Jennelle et al. [10] demonstrated how deer harvestingcould be implemented to manage chronic wasting disease prevalence in deer

http://crossmark.crossref.org/dialog/?doi=10.1098/rspb.2016.0952&domain=pdf&date_stamp=2016-09-28mailto:[email protected]://dx.doi.org/10.6084/m9.figshare.c.3469740https://dx.doi.org/10.6084/m9.figshare.c.3469740http://orcid.org/http://orcid.org/0000-0001-6983-6250http://orcid.org/0000-0002-0336-9706http://orcid.org/0000-0002-4801-9312http://creativecommons.org/licenses/by/4.0/http://creativecommons.org/licenses/by/4.0/http://creativecommons.org/licenses/by/4.0/http://rspb.royalsocietypublishing.org/

rspb.royalsocietypublishing.orgProc.R.Soc.B

283:20160952

2


populations. However, the analysis of the dynamics ofwildlife disease spread and the application of molecular epi-demiological techniques to investigate them are increasingonly slowly [11].

One effective approach to reconstructing emergence is touse phylodynamic techniques [12]. These methods are mosteffective when genetic data are available for large samplesthat have been serially sampled at known locations and forpathogens with high mutation rates and large populationsizes (such as fast-evolving viruses with RNA genomes). Suit-able datasets are more frequently available for humandiseases, those posing a zoonotic risk and those of economicimportance. Unfortunately, for other diseases of wildlife, therequired knowledge of pathogen diversity and host suscepti-bility is usually lacking and the genetic patterns may lacksufficient resolution [13]. Here, we develop an alternativeapproach to reconstruct pathogen spread and test hypothesesrelating to drivers of emergence, which could be used as amodel for other wildlife diseases. We demonstrate how citi-zen science can be employed to generate large datasets thatfeed spatio-temporal models of emergence, and how thisapproach can be integrated with the type of patchy geneticsampling that may be common to studies of wildlife diseases.

To study emergence of an infectious disease of amphi-bians (ranavirosis), we made use of an ongoing citizenscience project in the UK that has collated records of amphi-bian mortality for two decades and has provided material forgenetic characterization of the viruses responsible (genusRanavirus). Ranaviruses are large, double-stranded DNAviruses that infect and cause severe disease in amphibians,reptiles, and fish on five continents [14]. In the UK, ranavirusinfections and mass mortality events have been recorded andinvestigated since 1992 following alarm by members of thepublic [15]. Infections and reports of mortality have focusedon Rana temporaria a species which has shown medianvirus-driven declines of 81% sustained over a 12-yearperiod [15,16].

There is prima facie evidence linking ranavirus spread tohumans in a number of ways. Ranaviruses have been foundin traded amphibians [17], a number of outbreaks havebeen associated with introduced or farmed species [1821],and an earlier study that used microscopy and molecularmethods to compare viruses suggested that ranaviruseswere introduced to the UK from North America [22].Human activity has also been correlated with increased rana-virus prevalence in North America [23] and urbanization iscorrelated with ranavirus occurrence in the UK [24]. InNorth America, the use of infected juvenile salamanders asbait is known to have contributed to the spread of ranavirusin ambystomatid salamanders [25,26], which appears to rep-resent one incidence in a long history of human introductionsof pathogens to naive populations [27]. Here, we evaluate therelative importance of human-mediated spread versus otherpossible explanations for the apparent rapid recent spreadof ranavirus within the UK.

There is now a considerable body of research on climatechange as a driver of EIDs with documented effects on bothhosts and pathogens [28,29]. Climate change can alter amphi-bian behaviour such as timing and duration of hibernation,which may affect pathogen transmission opportunities[30,31]. Ranaviruses exhibit temperature sensitivity both inthe wild and in the laboratory. Ranavirus replication is morerapid at higher temperatures when grown under controlled

conditions in cell culture [32]. In an animal model, experimen-tally infected common frog tadpoles experience highermortality rates at higher temperature [33]. In the wild in theUK, ranavirus outbreaks show seasonality with a summerpeak [34]. Although it is problematic to extrapolate directlyfrom laboratory studies to ecology in the wild, such resultsindicate how climate change could alter the spatial distributionof ranavirosis.

Mapping of suspected ranavirosis events has consistentlyyielded a picture of apparent spread across England but hasnot previously accounted for reporting effort or consideredother potential biases in the data. We address these problemsand reconstruct spread using epidemiological models toassess whether classical epidemic spread, spatio-temporalpatterns in an environmental variable predicted to affecthostpathogen interactions, or human behaviour better pre-dict the emergence of the disease. We then combinedthis analysis with complementary information from an analy-sis of pathogen genotypes to reconstruct the pattern ofranavirus emergence.

2. Material and methods(a) Citizen science surveillance: The Frog Mortality

ProjectThe Frog Mortality Project (FMP) collated reports of amphibianmortality from the public between 1992 and 2013 before it wassubsumed into the Garden Wildlife Health project [35]. Methodsused to seek and administer reports changed somewhat over thisperiod (full details are provided in the electronic supplementarymaterial). Steps taken as part of this study to prepare the FMPrelational database for downstream analyses (particularly thegeoreference and time data) are also detailed in the electronicsupplementary material.

The reports were filtered for consistency with ranavirus infec-tion. Ulceration, red spots on the body, and limb necrosis/lossof digits were the signs of disease chosen to reliably representranavirosis [15,34]. Cunningham [34] showed a strong associationbetween these signs of disease (as recorded in reports of citizenscientists) and additional signs of disease at autopsy as well asthe presence of ranavirus in affected tissues [34]. From 1992 to1996, 95 carcases were examined from 24 sites of wild amphibianmortality at which lesions consistent with ranavirosis werereported, and 19 carcases from three sites with no such reports.Ranavirus was detected in at least one carcase (using virus culture)from 23 of the 24 sites with lesions reported, but none of the others.

A similar approach to filtering the FMP database for rana-virus-consistent mortality events has previously shown thatreporters observations of these signs can be used as a reliablepredictor of ranavirus occurrence [16]. As well as reports oflesions, we required mortality events to include at least five ani-mals, in order to be classified as consistent with ranavirusinfection. This rule makes use of the known virulence and infec-tivity of the virus [15,16] and replaced summer incidence, whichhad been used as a criterion by Teacher et al. [16]. This changewas made because recent studies identified incidents of rana-virus aetiology (confirmed by molecular methods) outsidesummerbetween March and October [36]. All remainingreports in the database were classed as negative.

(b) Covariate dataThe values of covariates were obtained at the resolution of dis-tricts, boroughs, and unitary authorities for England and Wales.Ordnance Survey Boundary Line data were obtained from the

http://rspb.royalsocietypublishing.org/


283:20160952

3


Edina Digimap service under OS OpenData license [37]. Monthlyaverages of climatic variables for all UK 5 5 km grid squarescovering the study period (19912010) were downloaded fromthe Met Office UKCP09 dataset. Regional human population den-sities were obtained from the Population Estimates Unit of theOffice for National Statistics. Covariates were decomposed byyear and by region (additional details of these methods are pro-vided in the electronic supplementary materials). Climaticvariables were strongly correlated (Pearsons correlation coeffi-cients range from 0.46 to 0.92, see the electronic supplementarymaterial, figure S1). We considered mean daily maximum temp-erature the most suitable climatic covariate, given the apparentpeak incidence of ranavirosis in summertime, and correlationsbetween temperature and virus growth in cell culture. Temporaland spatial patterns in population density and maximum temp-erature were visualized and analysed using linear regression in R.

(c) Two-component spatio-temporal modelsWe used twinstim, a function in the R package Surveillance v. 1.7[3840], to analyse the UK spread of ranavirus-consistent mor-tality events. Outbreaks were modelled as Poisson events. Theconditional intensity function (CIF) is the instantaneous rate orhazard for events at time, t, and location, s, conditioned on thehistory of all observations up to time, t [39]. The CIF is formu-lated as the sum of two componentsthe endemic andepidemic components (h() and e*() in the following equation):

lt, s ht, s et, s, 2:1

where l() is the function specifying a Poisson rate of infection.The definitions of the terms endemic and epidemic differfrom classical epidemiological definitions.

The epidemic component (function e*()) indicates the contri-bution to the infection rate due to transmission from existingoutbreaks and can be thought of as spread from pond to pondmediated via amphibian dispersal. This is sometimes termedthe self-exciting component [38] and describes the infectionpressure at a given time and location due to all other events inthe history up to that point. Interaction functions model thedecay of infectivity with distance (spatial interaction function,in the terminology of the R package) and time (temporalinteraction function) from the infection source [40].

The endemic component (function h()) is used to character-ize infections arising from sources outside of a conventionalsystem of transmission; i.e. they do not originate from a historicinfection but emergeare importedfrom outside of the trans-mission system. The function includes an offset, which we usedto allow for the number of amphibian ponds at risk, having con-trolled for reporting effort (see Controlling for reporting effort:estimating the at risk population)such that the endemic rateof infection is proportional to the relative number of pondsunder surveillance occupied by susceptible amphibians.

We explored the evidence for two alternative hypotheses forthe varying incidence of these endemic cases of ranavirosis:human translocation of virus, modelled by using human popu-lation density as a covariate, and climate-change effects,modelled using temperature as a covariate (daily maximumtemperature averaged across a calendar year).

(d) Model parametrizationUpper limits for the infectivity of events were set based on thebiology of frogs. We set the spatial limit for any pond to transmitinfection by the movement of infected individuals at 30 km(based on [41]) and the temporal limit at 2 920 days (an estimateof the maximum lifespan of a wild common frog). We usedhuman population density and daily maximum temperatureaveraged across a year as the variables in formulating the ende-mic components in the two competing models of spread.

To benchmark the performance of our final model, we generated500 unique covariate datasets by repeatedly randomizing regionto the remaining data and ran the two-component populationdensity model with these datasets as input.

Goodness of fit was assessed using the method of Ogata andimplemented as part of the Surveillance R package [39,42]. Wealso evaluated the model using simulated outbreaks based onthe fitted parameter values. We ran 100 simulations from thefitted model without providing any data as pre-history andsimulations with some pre-history. We compared the meantotal number of events and their spatial distribution to the realdata. The real counts were assessed against the 2.5% and97.5% quantiles of 100 realizations of the simulated model foreach region [39].

(e) Controlling for reporting effort: estimating the atrisk population

Reporting effort (number of citizen scientists recruited and therecords they generated) varied across years and regions. Inaddition, the density of populations at risk (ponds used by sus-ceptible common frog populations) also varied between regions.We reasoned that all of these issues could be allowed for by usingthe number of reports of mortality events that were negative (notranavirus consistent; mapped in figure 1). The number of theserecords would increase in proportion to the reporting effort,and would be proportional to the number of ponds in any onelocality. Their number, Nn, was therefore included as an offset(log-transformed since the Poisson model of events has a loglink-function).

It is possible that some reporting biases are not compensatedfor in this manner, for example, if the relative rates of positiveand negative reports were altered by the filtering of reportersor by appeals in the media to solicit public participation. Onesuch large drive took place in London and the South-East inthe early 1990s, and there have been other local and nationalmedia campaigns [34]. Of particular concern is the possibilitythat any association of outbreak incidence with human popu-lation density is actually a reflection of reporting bias, whichhas not been captured by our offset, since areas of high humanpopulation density are likely to have more reports. This issuewas investigated by noting that a reporting bias effect wouldaffect both endemic and epidemic components of the model,whereas long-distance translocations by humans (with an inci-dence proportional to human population density) would onlyaffect the endemic component. We therefore compared themodel with human population density included as a covariatein the endemic component with one in which it was includedin both components.

( f ) PhylogeneticsNucleotide sequences downloaded from the National Center forBiotechnology Information (NCBI) nucleotide database (listed inthe legend of figure 5) were aligned to sequence data from sevenUK ranaviruses [43], using BLAST to pull out homologousregions. Sequence alignment and phylogenetic tree constructionfollowed Price et al. [8]. There are no DNA sequence-derived esti-mates of substitution rates for ranaviruses, but rates acrossdsDNA viruses are thought to range from 1025 to 1028 substi-tutions per site per year (e.g. [44]). We used the upper limit tocalculate a maximum-likelihood estimate of the minimum timeto the most recent common ancestor of the UK viruses (Rscript provided; see the electronic supplementary material,Appendix S1). Support limits were calculated by taking valuescorresponding to two log-likelihood units either side of themaximum-likelihood estimate [45].


1992 positives 1997 positives 2002 positives 2010 positives

1992 negatives 1997 negatives 2002 negatives 2010 negatives

(a) (b) (c) (d)

(e) ( f ) (g) (h)

Figure 1. Visualization of UK ranavirus-consistent mortality events in time (1992 2010) and space (a d) and non-consistent frog mortality reports for the sameperiod (e h). (Online version in colour.)


283:20160952

4


3. ResultsAfter purging the citizen science database of records withessential data missing and obvious errors, a total of 4 460reports remained. Filtering the database for reports consistentwith ranavirosis in England and Wales produced a positiveset of 1 446 (32% of the total). Report numbersboth positiveand totalwere concentrated in particular years (see the elec-tronic supplementary material, figure S2) and regions (e.g.11% of total reports were received in 1995 from southeast Eng-land). Report data are visualized in space in figure 1, whichshows a time series of the changing distribution of reportsboth those consistent with ranavirosis (positive) and thosethat are not (negative). Both types of report accumulated overtime and their geographical distribution increased, althoughthe pattern was broadly similar for both types. It is not clearfrom these figures whether ranavirosis has spread or whetherreporting effort has driven the change in distribution.

(a) Spatial and temporal variation in main covariatesWhen we examined regional patterns in the variables associ-ated with our hypotheses (human population density andtemperature) we revealed some correlations, which can bevisualized on maps (figure 2). For example, London wasboth warmer and much more densely populated than otherregions, whilst Wales and parts of northern England werecooler and more sparsely populated. However, there were suf-ficient differences to discriminate between the two datasets; inparticular, temperature decreased in a fairly consistent wave-

like fashion from southeast England to Wales and northernEngland, whereas variation in population density was moreof a mosaic. In addition, there were differences in the trendsover time: mean daily maximum temperature across yearsincreased in nearly all study regions, with the majority increas-ing by 0.50.88C (inter-quartile range was 0.580.738C;electronic supplementary material, figure S3), and there wasa larger degree of change in the cooler north. Upward trendsin population density showed more variation between regions(electronic supplementary material, figure S3) with a greaterdegree of change in the south.

(b) Model outputs and performanceHuman population density models consistently outper-formed models with regional temperature as a covariate.The human population density models had higher likelihoodand lower Akaike information criterion (AIC) scores than thetemperature models (table 1). When fitting simple endemicmodels (excluding pond-to-pond infection), both these cov-ariates were significant terms. For more biologically realisticself-exciting models (including pond-to-pond infectionsvia amphibian dispersal), temperature was no longer a sig-nificant term in models that also contained populationdensity ( p 0.26). The model including only human popu-lation density (AIC 33 072, logLik 216 529), withdispersal between ponds modelled with a power-law func-tion, performed better than the equivalent temperature onlymodel (AIC 33 279, logLik 216 633). Human popula-tion density was a highly significant term ( p , 2 10216).


people per km2 mean temp. (C)

5 000

15

(a) (b)

Figure 2. Regional variation in main covariates averaged across study period, 1991 2010 (a) human population density ( people per square kilometre) and(b) mean maximum daily temperature (8C).

Table 1. Spatio-temporal model summaries (two-component models with power-law spatial interaction function or endemic component only) for each of theendemic covariates. All models include the number of negative records (see text) as an offset to control for reporting effort and represent the population atrisk.

model classlog-likelihood AIC endemic covariate p-value coefficient

two-component

population density 216 529 33 072 pop. density ,2 10216 4.89 1024

temperature 216 633 33 279 av. max. temp 0.97 4.17 1023

population density temperature 216 526 33 068 pop. densityav. max. temp

,2 10216

0.26

4.47 1024

9.95 1022

population density in both

components

216 529 33 074 endemic pop. density

epidemic pop. density

,2 10216

0.87

4.95 1024

24.58 1026

population density free schoolmeals

216 452 32 922 pop. density

free school meals

pop. density free schoolmeals

,2 10216

,0.0005

6 10214

1.74 1023

1.33 1021

29.23 1025

endemic only

population density 218 952 37 910 pop. density ,2 10216 4.58 1024

temperature 219 138 38 283 av. max. temp ,2 10216 9.92 1021

population density temperature 218 766 37 540 pop. densityav. max. temp

,2 10216

,2 102163.70 1024

6.47 1021

population density free schoolmeals

218 689 37 388 pop. density

free school meals

pop. density free schoolmeals

,2 10216

7.81 1029

,2 10216

1.31 1023

6.69 1022

25.85 1025


283:20160952

5


By contrast, in the comparable temperature model, tempera-ture was a non-significant term ( p 0.97). When humanpopulation density was included in both endemic andepidemic componentsas a test of whether the correlationwith human density was a reporting effort effectmodel performance was not improved (AIC 33 074,logLik 216 529) and population density on the epidemicside was a non-significant term ( p 0.87).

In both population density and temperature models, theendemic component which models the occurrence ofimported eventsexplained most events at the outset ofthe time series. Once such initial infections were established,

the estimates of transmission from pre-existing infections pre-dominated: in the two-component models the endemicproportion fell below 20% within 5 years and remainedfairly stable from then on. The estimates of infectivitydeclined rapidly with distance, with low rates over distancesin excess of 2 km. The power-law dispersal functions werealmost identical between temperature and population densitymodels. Both models had a residual distribution consistentwith the fitted Poisson CIF (see Material and methods andelectronic supplementary material, figure S4).

The population density model with the real data as inputperformed very well (lower AIC and higher log-likelihood


AIC

freq

uenc

y

32 900 33 100 33 300 33 500

0

5

10

15

20

25

30

popd

en

clim

ate

popd

en*f

sm

Figure 3. Comparison of model performance assessed by Akaike informationcriterion (AIC). Models featuring main covariatespopulation density (verti-cal line marked popden) and mean maximum daily temperature (verticalline marked climate)are compared with the extension of the populationdensity model including the interaction with the proportion of school stu-dents receiving free school meals (vertical line marked popden*fsm). Allmodels are compared to 500 iterations of the population density modelwhere each iteration used a unique randomization of region to the populationdensity data as input (bars of histogram). (Online version in colour.)

no. events

0156101115162021252630>30

Figure 4. Comparison of spatial point pattern for real versus simulated datafor 100 simulations from the fitted population density model with no dataprovided as pre-history. Intensity of shading represents the number of obser-vations of ranavirus outbreaks in the region. Triangles indicate regions wheresimulations overestimated (red triangle points up) or underestimated (bluetriangle points down) the real data. Regions where the real data fallinside 95% range of 100 realizations of the simulated model have no triangle.(Online version in colour.)


283:20160952

6


scores) when benchmarked against 500 iterations usingunique input datasets where region was repeatedly random-ized to the remaining data (figure 3). By contrast,the temperature model performance was similar to the ran-domized datasets. We also ran simulations from thetwo-component population density model as anothermeasure of its performance. The total number of events inthe real data was 1 446 and simulations did well in matchingthis with a mean of 1 538 without any data being provided aspre-history of outbreak locations. The model also performedwell in predicting where new events occurred, with thenumber of simulated events matching the real data well formost regions (figure 4). Exceptions were the southeast andthe northwest where, in the absence of pre-history data,simulations underestimated numbers given the high numbersof reports originating from these areas in the early years ofdata collection.

In London, the simulations predicted more events thanwere actually observed, which may be a consequence of themuch higher proportion of people living in apartmentblocks (with less access to ponds) compared with otherregions of the UK (UK Census Data, Office for NationalStatistics). To explore this hypothesis, we extended thetwo-component population density model to include aninteraction term between population density and the regionalproportion of school students receiving free school meals(a widely used proxy measure of socio-economic status inthe UK). We hypothesized that this variable would behighly correlated with the proportion of people withoutaccess to a garden and negatively correlated with the overallamount of green space in a region. We found that theinclusion of this interaction did indeed improve the modelfit (AIC 32 922, logLik 216 452; figure 3 and table 1) aswell as the match between the number of real and simula-ted events (1 446 and 1 505, respectively) and the number of

regions where simulations matched the real data (electronic sup-plementary material, figure S5).

(c) UK Ranavirus diversity revealed by virusphylogenetics

Our final multiple sequence alignment contained 2 267 basepairs from 23 virus isolates (seven from the UK and 16viruses from elsewhere), which we used to reconstruct a rana-virus phylogeny. The overall topology inferred by bothBayesian and maximum-likelihood methods was identical.UK viruses formed two groups with RUK13 and BUK3 form-ing an outgroup clade (figure 5). Monophyly of all UKranaviruses was not supported. Time to the most recentcommon ancestor of UK viruses (the node marked with ared star in figure 5) was estimated at 332 years ago (95%CI 189533 years ago) assuming a substitution rate of1025 subs site21 yr21.

4. DiscussionWe used data generated by a citizen science surveillance pro-ject in combination with occasional genetic sampling toreconstruct emergence of an important wildlife pathogen.By controlling for reporting effort and applying spatio-temporal modelling techniques, we have overcome thelimitations of common epidemiological techniques, such ascluster analyses. These approaches have dealt with our con-cern that the apparent geographical spread of UK ranavirusevents might be an artefact of reporting effort.

The use of an epidemic component in models and ourfinding that a high proportion of events were attributed toit showed that the majority of reported incidences of rana-virosis are likely to have arisen via transmission fromnearby ponds and we recovered estimates of dispersal con-sistent with the known ecology of frogs. This type of viraltransmission would have created a classical wave-likespread, with a timescale and spatial pattern that explainsmuch of the observed data. In addition, there was a small


0.0030

RUK128

CMTV_Aliva_Rt

BUK4

BNV_Tm

TFV

ADRV

STIV

CMTV

ESV

RUK216

RUK117

BUK3

ATV

FV3

BNV_Lb

RUK13

RGV

RUK126

EHNV

AAOV

CMTV_Aliva_Ao

CMTV_Mon

BNV_Nm

99

6766

100

93

100

98

98

85

85

100

89

100

4986

96

100

100

100

100

100100

100

100

100

100

72

51

9673

100

62

Figure 5. UK Ranavirus diversity in a global context. Monophyly of UK ranaviruses requires inclusion of Chinese and North American viruses. The tree was con-structed from seven concatenated multiple sequence alignments [8]. Node support values are annotated on the Bayesian tree and calculated using maximumlikelihood (bootstraps, bottom) and Bayesian inference ( posterior probabilities, top) under a GTR model of molecular evolution. Scale of branch lengths is in nucleo-tide substitutions/site. UK viruses are labelled in blue and labels start RUK or BUK. Additional sequences included are Frog virus 3 (FV3, AY548484), Tiger frog virus(TFV, AF389451), Ambystoma tigrinum virus (ATV, AY150217), Epizootic hematopoietic necrosis virus (EHNV, FJ433873), Soft-shelled turtle iridovirus (STIV, NC012637),Rana grylio virus (RGV, JQ654586), European sheatfish virus (ESV, JQ724856), Chinese giant salamander virus (ADRV, KC865735), Common midwife toad virus (CMTV,JQ231222), Boscas newt virus (accession numbers for individual loci as [8]). (Online version in colour.)


283:20160952

7


but significant proportion of events that were explained byendemic processes which model other sources of infectionincluding infection from non-local sources. We explicitly con-trolled for reporting effort by including the number of reportsof frog mortality not consistent with ranavirosis as an offsetin our models. In this way, the analysis answered the ques-tion where are the ponds with human observers? andforced the infection rate to be proportionate to this variable.Over and above this observer bias, the pattern of new out-breaks was strongly predicted by human populationdensity; we have interpreted this pattern as evidence forthe translocation of infectious materials by people, enhancingthe spread of a novel pathogen over greater geographical dis-tances at shorter timescales than could be accomplishedthrough typical frog movements.

The modelling process was correlational so requires theusual caveats of such studiesit is not possible to completelyrule out some other influence of human population densityon the outcomefor example, environmental pollutants

could have amplified the effects of pre-existing ranavirusinfections that had previously gone undetected. However,such hypotheses would require the virus to have been wide-spread already. Since we have shown that the majority ofrecorded events can be explained via transmission betweennearby ponds over the previous two decades, human translo-cations of infectious materials over a similar time periodseems a more parsimonious explanation for the endemiccontribution to the spread. The further improvement of thepopulation density model following inclusion of the inter-action with a measure of socio-economic status also addssupport to this interpretation: there is a predictable effect ofthis additional covariate on access to ponds but it is moredifficult to envisage how this interaction would modify acorrelation between human population density and thedetection of disease.

We used phylogenetic analysis based on some limitedsampling of infected tissues as a complementary approachto the spatio-temporal models and found support for



283:20160952

8


dispersal through human translocations when interpreted incombination with the modelling. Using our conservative esti-mate of the minimum time to the most recent commonancestor, it is clear that the genotypic diversity in UK virusescannot have arisen during the course of its spread over thelast 2530 years.

Hyatt et al. [22] previously obtained phylogenetic datasuggesting an introduction of ranavirus to the UK fromNorth America, possibly via the pet trade. In this context,our new phylogeny suggests that there have been at least twointroductions, each with a distinct history. It is likely thatfurther analysis of samples taken across the geographical distri-bution in figure 1 would identify other translocations includinglong-distance transfers within the UK that facilitated emer-gence. Previous work has identified several possible sourcesof such ranavirus introductions, by polymerase chain reaction(PCR) screening of animals that are traded, cultured, and inva-sive (e.g. North American bullfrogs, which have escaped fromfarms and the pet trade) [1721,46,47]. Human translocationsof infected animals have driven ranavirus emergence on abroad scale, for example, through the use of infectedsalamanders as angling bait in North America [25,26]. Preva-lence of ranavirus infection is associated with human activityin Canada and previous work has shown occurrence of rana-virosis in the UK to be associated with urban environments[23,24]. Some of the international spread of ranavirus may beassociated with the global trade in animals [17,48]. This tradeis huge in magnitude: for example, nearly 38 million animalsfrom 163 countries were imported to the USA in a 5-yearperiod at the turn of the century and 51 species of non-nativeamphibians and reptiles have been recorded in GreaterLondon since the 1980s [4951].

5. ConclusionOur results suggest further lines of research to help controlthe spread of ranavirus infections in the UK. Daszak et al.[9] identified two broad categories of human interventionaffecting the emergence of infectious diseases of wildlifewhich should be investigated: spread (i) by spillover of infec-tion from domestic animals and (ii) by human translocationsof pathogen or host. The first could be corroborated by theidentification of a vector; fish or non-native amphibians(e.g. North American bullfrog) being candidates for reservoirhosts [19,52]. The second category could involve the trans-location of fomites, such as aquatic plants, or infectedanimals, e.g. spawn, tadpoles, or frogs. Targeted samplingof such potential vectors, plus further genetic sampling ofranaviruses to gain a more complete picture of pathogendiversity would further address the mode and scale of trans-locations. In the meantime, existing recommendationsdiscouraging the movement of vectors and fomites could bemuch better publicized as an interim step.

This study also represents an important general contri-bution to the field of emerging wildlife disease throughthe demonstration of the potential and applicability ofits methodological approach. Our methods have enabledreconstruction of ongoing disease emergence in a timescaleenabling the information to flow into management decisions.This approach can be more widely useful when working witha pathogen where the mutation rate, biology, and practical-ities of sampling reduce the utility of fashionablephylodynamic techniques, which are more appropriate forfast evolving and intensively sampled RNA viruses. Emer-gent disease risks are posed by all types of pathogen, manyof which, like ranaviruses (DNA viruses), likely have lowermutation rates.

Although awareness of the ongoing biodiversity crisis hasincreased and is a clear and strong motivation for assemblingcomprehensive datasets, wildlife disease remains poorly rep-resented compared with disease affecting humans anddomestic animals. The approach used here, which builds ona citizen science surveillance project in combination withmainly opportunistic genetic sampling, therefore, representsa promising approach for the reconstruction of emergingwildlife diseases and exploration of hypotheses that caninform conservation strategies. To this end, we hope thisstudy will encourage others to both generate additional data-sets of this type (following initiatives like the Garden WildlifeHealth project, http://www.gardenwildlifehealth.org) and toapply the same approach to existing data.

Data accessibility. Citizen science database used in models and nucleotidesequence alignment used in phylogeny construction are available inData Dryad: http://dx.doi.org/10.5061/dryad.27c8s [43].Authors contributions. A.A.C. and T.E.S.L. established the Frog MortalityProject and UK Ranavirus isolate archive and organized the collec-tion of much of the report data. S.J.P., R.A.N., and T.W.J.G.designed the study. S.J.P. prepared the FMP database for analysis.S.J.P. and R.A.N. constructed and ran models and interpretedoutput. S.J.P. generated sequence data, conducted phylogeneticanalyses, and interpreted these analyses with R.A.N. S.J.P. wrotethe first draft of the manuscript and all authors contributedsubstantially to revisions.Competing interests. We have no competing interests.Funding. This work was supported by Natural Environment ResearchCouncil grant no. (NE/G011885/1, NE/M00080X/1, NE/M000338/1 and NE/M000591/1), the Systematics and Taxonomy (SynTax)research scheme administered by the Linnean Society of Londonand European Research Council grant 260801-BIG-IDEA.Acknowledgements. We thank Amanda Duffus for access to virus iso-lates, Sebastian Meyer for ongoing advice and support in applyingsurveillance functions to our data, and Rob Knell and two anon-ymous reviewers for helpful comments during manuscriptpreparation. The Frog Mortality Project was established by the Zool-ogical Society of London and Herpetofauna ConsultantsInternational and was largely administered by Froglife (registeredcharity number 1093372 in England & Wales).

References

1. Morens DM, Folkers GK, Fauci AS. 2004 Thechallenge of emerging and re-emerging infectiousdiseases. Nature 430, 242 249. (doi:10.1038/nature02759)

2. Jones KE, Patel NG, Levy MA, Storeygard A, Balk D,Gittleman JL, Daszak P. 2008 Global trends inemerging infectious diseases. Nature 451,990 993. (doi:10.1038/nature06536)

3. Barrett T, Rossiter PB. 1999 Rinderpest: the diseaseand its impact on humans and animals. Adv. VirusRes. 53, 89 110. (doi:10.1016/S0065-3527(08)60344-9)

http://www.gardenwildlifehealth.orghttp://www.gardenwildlifehealth.orghttp://dx.doi.org/10.5061/dryad.27c8shttp://dx.doi.org/10.5061/dryad.27c8shttp://dx.doi.org/10.1038/nature02759http://dx.doi.org/10.1038/nature02759http://dx.doi.org/10.1038/nature06536http://dx.doi.org/10.1016/S0065-3527(08)60344-9http://dx.doi.org/10.1016/S0065-3527(08)60344-9http://rspb.royalsocietypublishing.org/


283:20160952

9


4. Woodroffe R et al. 2008 Effects of culling on badgerabundance: implications for tuberculosis control.J. Zool. 274, 28 37.

5. LaDeau SL, Kilpatrick AM, Marra PP. 2007 West Nilevirus emergence and large-scale declines of NorthAmerican bird populations. Nature 447, 710 713.(doi:10.1038/nature05829)

6. Fisher MC, Garner TWJ, Walker SF. 2009 Globalemergence of Batrachochytrium dendrobatidis andAmphibian Chytridiomycosis in space, time, andhost. Annu. Rev. Microbiol. 63, 291 310. (doi:10.1146/annurev.micro.091208.073435)

7. Frick WF, Pollock JF, Hicks AC, Langwig KE, ReynoldsDS, Turner GG, Butchkoski CM, Kunz TH. 2010 Anemerging disease causes regional populationcollapse of a common North American bat species.Science 329, 679 682. (doi:10.1126/science.1188594)

8. Price SJ, Garner TWJ, Nichols RA, Balloux F, Ayres C,Mora-Cabello de Alba A, Bosch J. 2014 Collapse ofamphibian communities due to an introducedranavirus. Curr. Biol. 24, 2586 2591. (doi:10.1016/j.cub.2014.09.028)

9. Daszak P, Cunningham AA, Hyatt AD. 2000 Wildlifeecology - emerging infectious diseases of wildlife -threats to biodiversity and human health. Science287, 443 449. (doi:10.1126/science.287.5452.443)

10. Jennelle CS, Henaux V, Wasserberg G, Thiagarajan B,Rolley RE, Samuel MD. 2014 Transmission of chronicwasting disease in Wisconsin white-tailed deer:implications for disease spread and management.PLoS ONE 9, e91043. (doi:10.1371/journal.pone.0091043)

11. Benton CH, Delahay RJ, Trewby H, Hodgson DJ.2014 What has molecular epidemiology ever donefor wildlife disease research? Past contributions andfuture directions. Eur. J. Wildl. Res. 61, 1 16.(doi:10.1007/s10344-014-0882-4)

12. Grenfell BT et al. 2004 Unifying the epidemiologicaland evolutionary dynamics of pathogens. Science303, 327 332. (doi:10.1126/science.1090727)

13. MacPhee RDE, Greenwood AD. 2013 Infectiousdisease, endangerment, and extinction. Int. J. Evol.Biol. 16, 1 9. (doi:10.1155/2013/571939)

14. Gray MJ, Miller DL, Hoverman JT. 2009 Ecology andpathology of amphibian ranaviruses. Dis. Aquat.Organ. 87, 243 266. (doi:10.3354/dao02138)

15. Cunningham AA, Langton TES, Bennett PM, LewinJF, Drury SEN, Gough RE, Macgregor SK. 1996Pathological and microbiological findings fromincidents of unusual mortality of the common frog(Rana temporaria). Phil. Trans. R. Soc. Lond. B 351,1539 1557. (doi:10.1098/rstb.1996.0140)

16. Teacher AGF, Cunningham AA, Garner TWJ. 2010Assessing the long-term impact of Ranavirusinfection in wild common frog populations. Anim.Conserv. 13, 514 522. (doi:10.1111/j.1469-1795.2010.00373.x)

17. Schloegel LM, Picco AM, Kilpatrick AM, Davies AJ,Hyatt AD, Daszak P. 2009 Magnitude of the UStrade in amphibians and presence ofBatrachochytrium dendrobatidis and ranavirusinfection in imported North American bullfrogs

(Rana catesbeiana). Biol. Conserv. 142, 1420 1426.(doi:10.1016/j.biocon.2009.02.007)

18. Zhang QY, Xiao F, Li ZQ, Gui JF, Mao JH, ChincharVG. 2001 Characterization of an iridovirus from thecultured pig frog Rana grylio with lethal syndrome.Dis. Aquat. Organ. 48, 27 36. (doi:10.3354/dao048027)

19. Une Y, Sakuma A, Matsueda H, Nakai K, MurakamiM. 2009 Ranavirus outbreak in North AmericanBullfrogs (Rana catesbeiana), Japan, 2008. Emerg.Infect. Dis. 15, 1146 1147. (doi:10.3201/eid1507.081636)

20. Geng Y, Wang KY, Zhou ZY, Li CW, Wang J, He M,Yin ZQ, Lai WM. 2011 First report of a ranavirusassociated with morbidity and mortality in farmedchinese giant salamanders (Andrias davidianus).J. Comp. Pathol. 145, 95 102. (doi:10.1016/j.jcpa.2010.11.012)

21. Sharifian-Fard M, Pasmans F, Adriaensen C,Devisscher S, Adriaens T, Louette G, Martel A. 2011Ranavirosis in invasive bullfrogs, Belgium. Emerg.Infect. Dis. 17, 2371 2372. (doi:10.3201/eid1712.110236)

22. Hyatt AD, Gould AR, Zupanovic Z, Cunningham AA,Hengstberger S, Whittington RJ, Kattenbelt J,Coupar BEH. 2000 Comparative studies of piscineand amphibian iridoviruses. Arch. Virol. 145,301 331. (doi:10.1007/s007050050025)

23. St-Amour V, Wong WM, Garner TWJ, Lesbarreres D.2008 Anthropogenic influence on prevalence of 2amphibian pathogens. Emerg. Infect. Dis. 14,1175 1176. (doi:10.3201/eid1407.070602)

24. North AC, Hodgson DJ, Price SJ, Griffiths AGF. 2015Anthropogenic and ecological drivers of amphibiandisease (Ranavirosis). PLoS ONE 10, e0127037.(doi:10.1371/journal.pone.0127037)

25. Jancovich JK, Davidson EW, Parameswaran N, Mao J,Chinchar VG, Collins JP, Jacobs BL, Storfer A. 2005Evidence for emergence of an amphibian iridoviraldisease because of human-enhanced spread. Mol.Ecol. 14, 213 224. (doi:10.1111/j.1365-294X.2004.02387.x)

26. Picco AM, Collins JP. 2008 Amphibian commerce asa likely source of pathogen pollution. Conserv. Biol.22, 1582 1589. (doi:10.1111/j.1523-1739.2008.01025.x)

27. Cunningham AA, Daszak P, Rodriguez JP. 2003Pathogen pollution: defining a parasitological threatto biodiversity conservation. J. Parasitol. 89, S78 S83.

28. Harvell CD et al. 2002 Ecology - climate warmingand disease risks for terrestrial and marine biota.Science 296, 2158 2162. (doi:10.1126/science.1063699)

29. Altizer S, Ostfeld RS, Johnson PTJ, Kutz S, HarvellCD. 2013 Climate change and infectious diseases:from evidence to a predictive framework. Science341, 514 519. (doi:10.1126/science.1239401)

30. Garner TWJ, Rowcliffe JM, Fisher MC. 2011 Climatechange, chytridiomycosis or condition: anexperimental test of amphibian survival. Glob.Change Biol. 17, 667 675. (doi:10.1111/j.1365-2486.2010.02272.x)

31. Reading CJ. 1998 The effect of winter temperatureson the timing of breeding activity in the commontoad Bufo bufo. Oecologia 117, 469 475. (doi:10.1007/s004420050682)

32. Ariel E, Nicolajsen N, Christophersen M-B,Holopainen R, Tapiovaara H, Jensen BB. 2009Propagation and isolation of ranaviruses in cellculture. Aquaculture 294, 159 164. (doi:10.1016/j.aquaculture.2009.05.019)

33. Bayley AE, Hill BJ, Feist SW. 2013 Susceptibility ofthe European common frog Rana temporaria to apanel of ranavirus isolates from fish and amphibianhosts. Dis. Aquat. Organ. 103, 171 183. (doi:10.3354/dao02574)

34. Cunningham AA. 2001 Investigations into massmortalities of the common frog (Rana temporaria)in Britain: epidemiology and aetiology. RoyalVeterinary College (University of London) (cited 9May 2013). See http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.269010.

35. Lawson B, Petrovan SO, Cunningham AA. 2015Citizen science and wildlife disease surveillance.EcoHealth 12, 693 702. (doi:10.1007/s10393-015-1054-z)

36. Animal Health and Veterinary Laboratories Agency.2013 GB Wildlife Disease Surveillance Partnershipreports - Quarterly. 2013 (cited 13 February 2015).See http://webarchive.nationalarchives.gov.uk/20140707135733/http://www.defra.gov.uk/ahvla-en/publication/wildlife-survreports/.

37. Ordnanace Survey, GB. OS Boundary-Line Districts,boroughs and unitary authorities [Shapefilegeospatial data], November 2012 version. Coverage:Great Britain, Updated May 2015. Using: EDINADigimap Ordnance Survey Service, http://digimap.edina.ac.uk/. Downloaded: June 2015.

38. Meyer S. 2010 Spatio-temporal infectious diseaseepidemiology based on point processes [mast].Institut fur Statistik. See http://epub.ub.uni-muenchen.de/11703/.

39. Meyer S, Elias J, Hoehle M. 2012 A space-timeconditional intensity model for invasivemeningococcal disease occurrence. Biometrics 68,607 616. (doi:10.1111/j.1541-0420.2011.01684.x)

40. Meyer S, Held L. 2014 Power-law models forinfectious disease spread. Ann. Appl Stat. 8,1612 1639. (doi:10.1214/14-AOAS743)

41. Hitchings SP, Beebee TJC. 1997 Geneticsubstructuring as a result of barriers to gene flow inurban Rana temporaria (common frog) populations:implications for biodiversity conservation.Heredity 79, 117 127. (doi:10.1038/hdy.1997.134)

42. Ogata Y. 1988 Statistical-models for earthquakeoccurrences and residual analysis for point-processes. J. Am. Stat. Assoc. 83, 9 27. (doi:10.1080/01621459.1988.10478560)

43. Price SJ, Garner TWJ, Cunningham AA, Langton TES,Nichols RA. Data from: Reconstructing theemergence of a lethal infectious disease of wildlifesupports a key role for spread throughtranslocations by humans. Dryad Digital Repository.(http://dx.doi.org/10.5061/dryad.27c8s)

http://dx.doi.org/10.1038/nature05829http://dx.doi.org/10.1146/annurev.micro.091208.073435http://dx.doi.org/10.1146/annurev.micro.091208.073435http://dx.doi.org/10.1126/science.1188594http://dx.doi.org/10.1126/science.1188594http://dx.doi.org/10.1016/j.cub.2014.09.028http://dx.doi.org/10.1016/j.cub.2014.09.028http://dx.doi.org/10.1126/science.287.5452.443http://dx.doi.org/10.1371/journal.pone.0091043http://dx.doi.org/10.1371/journal.pone.0091043http://dx.doi.org/10.1007/s10344-014-0882-4http://dx.doi.org/10.1126/science.1090727http://dx.doi.org/10.1155/2013/571939http://dx.doi.org/10.3354/dao02138http://dx.doi.org/10.1098/rstb.1996.0140http://dx.doi.org/10.1111/j.1469-1795.2010.00373.xhttp://dx.doi.org/10.1111/j.1469-1795.2010.00373.xhttp://dx.doi.org/10.1016/j.biocon.2009.02.007http://dx.doi.org/10.3354/dao048027http://dx.doi.org/10.3354/dao048027http://dx.doi.org/10.3201/eid1507.081636http://dx.doi.org/10.3201/eid1507.081636http://dx.doi.org/10.1016/j.jcpa.2010.11.012http://dx.doi.org/10.1016/j.jcpa.2010.11.012http://dx.doi.org/10.3201/eid1712.110236http://dx.doi.org/10.3201/eid1712.110236http://dx.doi.org/10.1007/s007050050025http://dx.doi.org/10.3201/eid1407.070602http://dx.doi.org/10.1371/journal.pone.0127037http://dx.doi.org/10.1111/j.1365-294X.2004.02387.xhttp://dx.doi.org/10.1111/j.1365-294X.2004.02387.xhttp://dx.doi.org/10.1111/j.1523-1739.2008.01025.xhttp://dx.doi.org/10.1111/j.1523-1739.2008.01025.xhttp://dx.doi.org/10.1126/science.1063699http://dx.doi.org/10.1126/science.1063699http://dx.doi.org/10.1126/science.1239401http://dx.doi.org/10.1111/j.1365-2486.2010.02272.xhttp://dx.doi.org/10.1111/j.1365-2486.2010.02272.xhttp://dx.doi.org/10.1007/s004420050682http://dx.doi.org/10.1007/s004420050682http://dx.doi.org/10.1016/j.aquaculture.2009.05.019http://dx.doi.org/10.1016/j.aquaculture.2009.05.019http://dx.doi.org/10.3354/dao02574http://dx.doi.org/10.3354/dao02574http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.269010http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.269010http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.269010http://dx.doi.org/10.1007/s10393-015-1054-zhttp://dx.doi.org/10.1007/s10393-015-1054-zhttp://webarchive.nationalarchives.gov.uk/20140707135733/http://www.defra.gov.uk/ahvla-en/publication/wildlife-survreports/http://webarchive.nationalarchives.gov.uk/20140707135733/http://www.defra.gov.uk/ahvla-en/publication/wildlife-survreports/http://webarchive.nationalarchives.gov.uk/20140707135733/http://www.defra.gov.uk/ahvla-en/publication/wildlife-survreports/http://webarchive.nationalarchives.gov.uk/20140707135733/http://www.defra.gov.uk/ahvla-en/publication/wildlife-survreports/http://webarchive.nationalarchives.gov.uk/20140707135733/http://www.defra.gov.uk/ahvla-en/publication/wildlife-survreports/http://digimap.edina.ac.uk/http://digimap.edina.ac.uk/http://digimap.edina.ac.uk/http://epub.ub.uni-muenchen.de/11703/http://epub.ub.uni-muenchen.de/11703/http://epub.ub.uni-muenchen.de/11703/http://dx.doi.org/10.1111/j.1541-0420.2011.01684.xhttp://dx.doi.org/10.1214/14-AOAS743http://dx.doi.org/10.1038/hdy.1997.134http://dx.doi.org/10.1038/hdy.1997.134http://dx.doi.org/10.1080/01621459.1988.10478560http://dx.doi.org/10.1080/01621459.1988.10478560http://dx.doi.org/10.5061/dryad.27c8shttp://dx.doi.org/10.5061/dryad.27c8shttp://rspb.royalsocietypublishing.org/

rspb.royalsocietypublishing.orgProc.

10


44. Firth C, Kitchen A, Shapiro B, Suchard MA, HolmesEC, Rambaut A. 2010 Using time-structured data toestimate evolutionary rates of double-stranded DNAviruses. Mol. Biol. Evol. 27, 2038 2051. (doi:10.1093/molbev/msq088)

45. Wilks SS. 1938 The large-sample distribution of thelikelihood ratio for testing composite hypotheses. Ann.Math. Stat. 9, 60 62. (doi:10.1214/aoms/1177732360)

46. Schlaepfer MA, Hoover C, Dodd CK. 2005 Challengesin evaluating the impact of the trade in amphibiansand reptiles on wild populations. Bioscience 55,256 264. (doi:10.1641/0006-3568(2005)055[0256:CIETIO]2.0.CO;2)

47. Mazzoni R et al. 2009 Mass mortality associated with afrog virus 3-like Ranavirus infection in farmed tadpolesRana catesbeiana from Brazil. Dis. Aquat. Organ. 86,181 191. (doi:10.3354/dao02096)

48. DEFRA. 2012 Assessment of viral disease threats to wildpopulations of amphibians and fish in Great Britain -FC1194. CEFAS; 2012 (cited 13 March 2016). Report No.:FC1194. See http://goo.gl/meK8nh.

49. Langton TES, Herbert C. 2011 On the distribution,ecology and management of non-native reptilesand amphibians in the London Area. Part 2. Diseaseimpacts, perspectives, trade exploitation and findingethical solutions. Lond. Nat. 90, 157 177.

50. Langton TES, Atkins W, Herbert C. 2011 On thedistribution, ecology and management of non-native reptiles and amphibians in the London area.Part 1. Distribution and predatory/prey impacts.Lond. Nat. 90, 83 155.

51. Marano N, Arguin PM, Pappaioanou M. 2007 Impactof globalization and animal trade on infectiousdisease ecology. Emerg. Infect. Dis. 13, 1807 1809.(doi:10.3201/eid1312.071276)

52. Brenes R, Gray MJ, Waltzek TB, Wilkes RP, Miller DL.2014 Transmission of ranavirus between ectothermicvertebrate hosts. PLoS ONE 9, e92476. (doi:10.1371/journal.pone.0092476)

R

.So

c.B283:20160952

http://dx.doi.org/10.1093/molbev/msq088http://dx.doi.org/10.1093/molbev/msq088http://dx.doi.org/10.1214/aoms/1177732360http://dx.doi.org/10.1641/0006-3568(2005)055[0256:CIETIO]2.0.CO;2http://dx.doi.org/10.1641/0006-3568(2005)055[0256:CIETIO]2.0.CO;2http://dx.doi.org/10.3354/dao02096http://goo.gl/meK8nhhttp://goo.gl/meK8nhhttp://dx.doi.org/10.3201/eid1312.071276http://dx.doi.org/10.1371/journal.pone.0092476http://dx.doi.org/10.1371/journal.pone.0092476http://rspb.royalsocietypublishing.org/

Reconstructing the emergence of a lethal infectious disease of wildlife supports a key role for spread through translocations by humansBackgroundMaterial and methodsCitizen science surveillance: The Frog Mortality ProjectCovariate dataTwo-component spatio-temporal modelsModel parametrizationControlling for reporting effort: estimating the at risk populationPhylogenetics

ResultsSpatial and temporal variation in main covariatesModel outputs and performanceUK Ranavirus diversity revealed by virus phylogenetics

DiscussionConclusionData accessibilityAuthors contributionsCompeting interestsFundingAcknowledgementsReferences

Documents

Reconstructing the emergence of a lethal infectious ...rspb.royalsocietypublishing.org/content/royprsb/283/1839/20160952...Electronic supplementary material is available ... 3School